Bandwidth selection for the presmoothed logrank test
- M.A. Jácome Pumar 1
- I. López de Ullibarri 2
- 1 Facultade de Ciencias, Universidade da Coruña
- 2 Escuela Universitaria Polit´ecnica, Universidade da Coruña
- Ginzo Villamayor, María José (ed. lit.)
- Alonso Meijide, José María (ed. lit.)
- Ramil Novo, Luis Alberto (ed. lit.)
Editorial: Sociedade Galega para a Promoción da Estatística e da Investigación de Operacións (SGAPEIO) ; Servizo de Publicacións ; Deputación de Lugo
ISBN: 978-84-8192-522-7
Ano de publicación: 2015
Páxinas: 63-73
Congreso: Congreso galego de Estatística e Investigación de Operacións (12. 2015. Lugo)
Tipo: Achega congreso
Resumo
The most popular test in the two-sample random censorship model is the logrank test. It is based on the comparison of the Nelson-Aalen estimates of the corresponding cumulative hazard functions. A new logrank-type test, based on the presmoothed counterpart of the Nelson-Aalen estimators, has been shown to have the proper size under the null hypothesis, while improving the power over a wide range of alternatives. The success of the presmoothed methods depends on the choice of a smoothing parameter or bandwidth. We propose two data-driven bandwidth selectors that maximize the power while keeping the size of the test at a given level, and show via simulations the significant improvement of the presmoothed logrank test over the classical one. The application of the new test with both bandwidth selectors is illustrated with several real data examples